Episodic Memory
Continued Influence Task
Also called: continued influence effect task, continued influence paradigm, misinformation retraction task, correction of misinformation task, CIE paradigm
Measure the continued influence of misinformation: a detail in a short news story keeps shaping inferences after a later message corrects it.
9-12 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Try it here first, with no account.
- Papers, 2021 to 2025
- 97
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures the continued influence of misinformation: a detail keeps shaping inferences after it is corrected. Participants read six short news stories about invented events; a detail is later corrected in two, never corrected in two and never given in two. They then rate statements that follow from each detail. The main score is agreement in the corrected stories minus agreement in the stories without the detail. The last screen names the corrected details.
Constructs
- Continued influence of misinformation
- Memory updating
- Inferential reasoning
- Correction of beliefs
Research uses
- Misinformation and correction research
- Memory updating and reasoning research
- Within-study comparison of groups and conditions
How the paradigm works
The Continued Influence Task measures the continued influence of misinformation: a detail that explains an event keeps shaping how people reason about it after the detail has been corrected. Wilkes and Leatherbarrow (1988) found that inferences from a message later identified as wrong went on being drawn although people recalled the correction, and Johnson and Seifert (1994) showed it with a series of messages about a warehouse fire, in which an early message named a cause and a later one retracted it. This version follows the within-person form of Ecker and Antonio (2021, Experiment 2): participants read six short news stories about invented events in a small town, each told in five messages shown one at a time. In two stories, message 2 gives a detail that explains the event and message 4 corrects it, saying only that the detail was wrong. In two, the detail is given and never corrected. In two, it is never given.
More about the paradigm
After the six stories and 60 seconds of true-or-false arithmetic (0 to 180 as a setting), participants answer questions about each story in reading order: three statements that follow from the detail, each rated from 1 (strongly disagree) to 6 (strongly agree); a question about a fact that every version of the story gives, with three answers; and whether a later message corrected an earlier one. As in McIlhiney and colleagues (2023), the first story is never corrected and no two corrected stories come in a row, so corrections do not become expected. The session's seed sets the order and which kind each story is, so every story is each kind in a third of sessions. The stories were written for this task about ordinary causes, such as a broken-down truck or a leaking roof, with nothing about health, politics, religion or groups of people, and they are written in each of the app's eight languages. The last screen says that the stories were invented and shows each corrected detail with its correction.
The primary score is the continued influence: the mean agreement in the corrected stories minus that in the stories without the detail, in scale points. Above 0 means the corrected detail still raised agreement. Scores also give the influence of the uncorrected detail, how far the correction lowered agreement, the continued influence over the corrected stories whose correction the participant reported noticing, and the fact questions. In Ecker and Antonio's (2021) Experiment 2, with 68 students rating statements from 0 to 10, mean inference scores were 3.13 without the detail and 4.93 with the detail uncorrected; after corrections from sources of high and low trustworthiness they were 5.24 and 5.88, so the corrections did not lower reliance. McIlhiney and colleagues (2023) found that a retraction lowered reliance but left it above zero in 255 students, and that a per-person difference over eight reports had an internal consistency of .57, so one person's score from six stories is noisy and is best read across a group. A second session would show the stories again after their corrections were named, so the test is meant to be taken once by each person, and a repeat session is flagged. Scores describe how corrections change inferences in this task, a research measure; they do not diagnose.
Scores
- Continued Influencepoints
- Mean agreement in the stories whose detail was corrected minus that in the stories without the detail, on the scale from 1 to 6.
- Above 0 means the corrected detail still raised agreement
- Uncorrected Influencepoints
- Mean agreement in the stories whose detail was never corrected minus that in the stories without the detail.
- How far the detail raised agreement when it stood
- Correction Reductionpoints
- Mean agreement in the uncorrected stories minus that in the corrected stories.
- Above 0 means the correction lowered agreement
- Agreement, Corrected Storiespoints
- Mean agreement with the statements of the stories whose detail was corrected, from 1 (strongly disagree) to 6 (strongly agree).
- Compare with the stories without the detail
- Agreement, Uncorrected Storiespoints
- Mean agreement with the statements of the stories whose detail was never corrected.
- Compare with the corrected stories
- Agreement, Stories Without the Detailpoints
- Mean agreement with the statements of the stories that never gave the detail.
- The baseline
- Continued Influence, Correction Noticedpoints
- The continued influence over only the corrected stories whose correction the participant reported noticing; empty when none was reported.
- Above 0 means the detail was used even when its correction was noticed
- Corrections Noticed%
- Share of corrected stories for which the participant answered Yes to the question about a correction.
- 0 is flagged for review
- Corrections Reported Wrongly%
- Share of the other stories for which the participant answered Yes to the question about a correction.
- Context for the correction question
- Fact Accuracy%
- Share of fact questions answered right; each has three answers.
- At or under one in three is flagged for review
- Statements Ratedcount
- Agreement ratings in the scored run.
- Fewer than 90% of those planned is invalid
- Statements Plannedcount
- Statements the session set out to rate.
- 18 in every session
- Scale Points Usedcount
- How many of the six scale points the ratings used.
- 1 means the same rating throughout, which is invalid
- Median Rating Timems
- Median time to rate a statement, leaving out ratings under 100 ms.
- Context for the ratings
- Ratings Under 100 mscount
- Ratings given within 100 ms of the statement. They still count as ratings.
- Expected to be 0
- Median Reading Timeseconds
- Median time on a message before pressing Next; the button wakes after 3 seconds.
- Context for the reading
- Delayseconds
- Seconds of arithmetic between the stories and the questions.
- Setting context
- Earlier Sessionscount
- This participant's earlier sessions of the task, when scoring counted them.
- Above 0 is flagged for review
Use and evidence
97 papers from 2021 to 2025 name this task in their title or abstract, up 106% compared with 2016 to 2020. That ranks it 219 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 25 |
| 2024 | 35 |
| 2023 | 14 |
| 2022 | 15 |
| 2021 | 8 |
| 2020 | 13 |
| 2019 | 11 |
| 2018 | 15 |
| 2017 | 6 |
| 2016 | 2 |
| 2015 | 6 |
| 2014 | 1 |
| 2013 | 1 |
| 2012 | 1 |
| 2011 | 1 |
| 2010 | 2 |
| 2009 | 0 |
| 2008 | 0 |
| 2007 | 0 |
| 2006 | 0 |
| 2005 | 0 |
| 2004 | 0 |
| 2003 | 0 |
| 2002 | 1 |
| 2001 | 0 |
| 2000 | 0 |
Use grade D
30 to 99 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 0
- Papers on reliability or validity
- 10
- Registered clinical studies using it as an outcome
- 0
How the grades work
Use: A is 1,000 or more papers from 2021 to 2025, B 300 to 999, C 100 to 299, D 30 to 99 and E fewer than 30. Evidence: A means at least 25 papers report its test-retest reliability and at least 25 registered clinical studies use it as an outcome; B, at least 8 test-retest papers, and either 5 registered clinical studies or 150 reliability or validity papers; C, at least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies; D, little published measurement evidence found. The counts measure use. They include every paper that names the task in its title or abstract. Grades describe the published research on this task, not ConductCognition's version of it.
Source: ConductCognition test census 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Stories
- 6 short news stories about invented events in a small town, written for this task, each told in 5 messages shown one at a time. No story is about health, politics, religion or groups of people. The stories are in the session's language.
- Kinds of story
- In every session, 2 stories give a detail in message 2 and correct it in message 4, 2 give it and never correct it, and 2 never give it. The first story is never corrected and no two corrected stories come in a row. The session's seed sets the order and which story is which kind.
- Reading
- Each message's Next button wakes after 3 seconds.
- Delay
- 60 seconds of true-or-false arithmetic by default, 0 to 180 as a setting.
- Questions
- For each story, in reading order: 3 statements that follow from the detail, each rated from 1 (strongly disagree) to 6 (strongly agree); a fact question with 3 answers; and whether a later message corrected an earlier one. No time limit.
- Last screen
- Says that the stories were invented and shows each corrected detail with its correction.
- Score
- Continued influence: the mean agreement in the corrected stories minus that in the stories without the detail, in scale points.
The stories, statements and questions were written for this task; no published scenario is used. A second session would repeat the stories after their corrections were named, so each person takes the task once, and a repeat session is flagged.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: continued_influence_inference from the scored run, one agreement rating from 1 to 6 per statement, untimed, with the fact and correction questions and the messages from the same run. A rating is a judgment with no correct option, so accuracy is not judged against chance. The same rating on every statement invalidates, and so do fewer than 90% of the planned statements rated. Ratings under 200 ms on at least half of 4 or more statements invalidate. Fact accuracy at or under chance (one in three) is flagged and withholds the percentile; no correction reported as noticed and an earlier session of this test are flagged for review. A hidden tab withholds the percentile.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. The same rating on every statement cannot depend on the stories. The score compares three kinds of story, two stories each, so it needs nearly all the ratings; the 90% limit is a ConductScience choice. Fact questions about details every version of a story shows check that the stories were read, as McIlhiney and colleagues (2023) excluded participants who failed most of theirs; a correction that was never noticed leaves a corrected story the same as an uncorrected one, as Connor Desai and Reimers (2019) showed by setting such participants apart. A second session has read every story and been told which details were corrected, so an earlier session is flagged. An answer under 200 ms is faster than a simple visual reaction, so it cannot reflect the item. Following response-time effort screening for untimed tests (Wise and Kong 2005), the session is invalid only when at least half the answers are that fast. A hidden tab lets the study, delay or response timers run without the participant, so a percentile is withheld after one.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Delay before the questions (seconds) | 60 | 0 to 180 |
Yearly plans can change settings within these limits. Free and pay-as-you-go studies use the standard settings.
Need a setting this test does not have?
Devices and languages
- Untimed test: runs on desktops, laptops, tablets and phones. Participant devices
- Languages: English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese.
- Available on: Every plan, including the free plan.
- No percentile. Scores and quality flags only. How norms work
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Continued Influence Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Wilkes & Leatherbarrow, 1988; Johnson & Seifert, 1994; Ecker & Antonio, 2021), with the standard settings (delay before the questions (seconds): 60). The main score was continued influence (mean agreement in the stories whose detail was corrected minus that in the stories without the detail, on the scale from 1 to 6).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Johnson, H. M., & Seifert, C. M. (1994). Sources of the continued influence effect: When misinformation in memory affects later inferences. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20(6), 1420-1436. https://doi.org/10.1037/0278-7393.20.6.1420
- Wilkes, A. L., & Leatherbarrow, M. (1988). Editing episodic memory following the identification of error. Quarterly Journal of Experimental Psychology Section A, 40(2), 361-387. https://doi.org/10.1080/02724988843000168
- Ecker, U. K. H., & Antonio, L. M. (2021). Can you believe it? An investigation into the impact of retraction source credibility on the continued influence effect. Memory & Cognition, 49(4), 631-644. https://doi.org/10.3758/s13421-020-01129-yFive-message reports in a within-person design with a control that never gives the detail.
- McIlhiney, P., Gignac, G. E., Ecker, U. K. H., Kennedy, B. L., & Weinborn, M. (2023). Executive function and the continued influence of misinformation: A latent-variable analysis. PLOS ONE, 18(4), e0283951. https://doi.org/10.1371/journal.pone.0283951Several stories per person, the first never corrected and corrections never in a row, with rated statements and fact questions.
- Connor Desai, S., & Reimers, S. (2019). Comparing the use of open and closed questions for Web-based measures of the continued-influence effect. Behavior Research Methods, 51(3), 1426-1440. https://doi.org/10.3758/s13428-018-1066-zClosed questions for measuring the effect online.
Source papers
- Johnson HM, Seifert CM (1994). Sources of the continued influence effect: when misinformation in memory affects later inferences. Journal of Experimental Psychology: Learning, Memory, and Cognition, 20(6):1420-1436.Paradigm source, Adults
- Ecker UKH, Antonio LM (2021). Can you believe it? An investigation into the impact of retraction source credibility on the continued influence effect. Memory & Cognition, 49(4):631-644.Paradigm source, Students, Australia
- McIlhiney P, Gignac GE, Ecker UKH, Kennedy BL, Weinborn M (2023). Executive function and the continued influence of misinformation: a latent-variable analysis. PLOS ONE, 18(4):e0283951.Paradigm source, Students, Australia
- Connor Desai S, Reimers S (2019). Comparing the use of open and closed questions for Web-based measures of the continued-influence effect. Behavior Research Methods, 51(3):1426-1440.Paradigm source, Adults, online
Questions
How long does the Continued Influence Task take?
About 9 to 12 minutes, as listed in the test library.
What device does a participant need?
Untimed test: runs on desktops, laptops, tablets and phones. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Continued Influence, Uncorrected Influence, Correction Reduction, Agreement, Corrected Stories, Agreement, Uncorrected Stories, Agreement, Stories Without the Detail, Continued Influence, Correction Noticed, Corrections Noticed, Corrections Reported Wrongly, Fact Accuracy, Statements Rated, Statements Planned, Scale Points Used, Median Rating Time, Ratings Under 100 ms, Median Reading Time, Delay and Earlier Sessions. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Continued Influence Task used in research?
97 papers from 2021 to 2025 name it in their title or abstract, up 106% compared with 2016 to 2020. That ranks it 219 of 495 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Continued Influence Task have?
In the published research, 0 papers report its test-retest reliability and 0 registered clinical studies use it as an outcome (evidence grade D in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Delay before the questions (seconds) (0 to 180). Settings can be changed on yearly plans; free and pay-as-you-go studies use the standard settings.
Which languages can participants use?
Participants can take this test in English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese, with its words in the same language. English is on every plan, and the other languages come with yearly plans. Have a native speaker read the machine-translated text before launch.
Is this a clinical test?
No. ConductCognition is for research use only. It runs, scores and exports the test; clinical interpretation is outside the platform.
Every test: the full test list.
Run this test in your next study
Start free with 10 free participants on any test, then use participant credits or a yearly plan.